39

1. Regular Expression Tutorial
In this tutorial, I will teach you all you need to know to be able to craft powerful time-saving regular expressions. I will start with the most basic concepts, so that you can follow this tutorial even if you know nothing at all about regular expressions yet. But I will not stop there. I will also explain how a regular expression engine works on the inside, and alert you at the consequences. This will help you to understand quickly why a particular regex does not do what you initially expected. It will save you lots of guesswork and head-scratching when you need to write more complex regexes.

What Regular Expressions Are Exactly - Terminology
Basically, a regular expression is a pattern describing a certain amount of text. Their name comes from the mathematical theory on which they are based. But we will not dig into that. Since most people including myself are lazy to type, you will usually find the name abbreviated to regex or regexp. I prefer regex, because it is easy to pronounce the plural "regexes". In this book, regular expressions are printed guillemots: «regex». They clearly separate the pattern from the surrounding text and punctuation. This first example is actually a perfectly valid regex. It is the most basic pattern, simply matching the literal text „regex”. A "match" is the piece of text, or sequence of bytes or characters that pattern was found to correspond to by the regex processing software. Matches are indicated by double quotation marks, with the left one at the base of the line. «\b[A-Z0-9._%-]+@[A-Z0-9._%-]+\.[A-Z0-9._%-]{2,4}\b» is a more complex pattern. It describes a series of letters, digits, dots, percentage signs and underscores, followed by an at sign, followed by another series of letters, digits, dots, percentage signs and underscores, finally followed by a single dot and between two and four letters. In other words: this pattern describes an email address. With the above regular expression pattern, you can search through a text file to find email addresses, or verify if a given string looks like an email address. In this tutorial, I will use the term "string" to indicate the text that I am applying the regular expression to. I will indicate strings using regular double quotes. The term “string” or “character string” is used by programmers to indicate a sequence of characters. In practice, you can use regular expressions with whatever data you can access using the application or programming language you are working with.

Different Regular Expression Engines
A regular expression “engine” is a piece of software that can process regular expressions, trying to match the pattern to the given string. Usually, the engine is part of a larger application and you do not access the engine directly. Rather, the application will invoke it for you when needed, making sure the right regular expression is applied to the right file or data. As usual in the software world, different regular expression engines are not fully compatible with each other. It is not possible to describe every kind of engine and regular expression syntax (or “flavor”) in this tutorial. I will focus on the regex flavor used by Perl 5, for the simple reason that this regex flavor is the most popular

40 one, and deservedly so. Many more recent regex engines are very similar, but not identical, to the one of Perl 5. Examples are the open source PCRE engine (used in many tools and languages like PHP), the .NET regular expression library, and the regular expression package included with version 1.4 and later of the Java JDK. I will point out to you whenever differences in regex flavors are important, and which features are specific to the Perl-derivatives mentioned above.

Give Regexes a First Try
You can easily try the following yourself in a text editor that supports regular expressions, such as EditPad Pro. If you do not have such an editor, you can download the free evaluation version of EditPad Pro to try this out. EditPad Pro's regex engine is fully functional in the demo version. As a quick test, copy and paste the text of this page into EditPad Pro. Then select Edit|Search and Replace from the menu. In the search pane that appears near the bottom, type in «regex» in the box labeled “Search Text”. Mark the “Regular expression” checkbox, unmark “All open documents” and mark “Start from beginning”. Then click the Search button and see how EditPad Pro's regex engine finds the first match. When “Start from beginning” is checked, EditPad Pro uses the entire file as the string to try to match the regex to. When the regex has been matched, EditPad Pro will automatically turn off “Start from beginning”. When you click the Search button again, the remainder of the file, after the highlighted match, is used as the string. When the regex can no longer match the remaining text, you will be notified, and “Start from beginning” is automatically turned on again. Now try to search using the regex «reg(ular expressions?|ex(p|es)?)». This regex will find all names, singular and plural, I have used on this page to say “regex”. If we only had plain text search, we would have needed 5 searches. With regexes, we need just one search. Regexes save you time when using a tool like EditPad Pro. If you are a programmer, your software will run faster since even a simple regex engine applying the above regex once will outperform a state of the art plain text search algorithm searching through the data five times. Regular expressions also reduce development time. With a regex engine, it takes only one line (e.g. in Perl, PHP, Java or .NET) or a couple of lines (e.g. in C using PCRE) of code to, say, check if the user's input looks like a valid email address.

41

2. Literal Characters
The most basic regular expression consists of a single literal character, e.g.: «a». It will match the first occurrence of that character in the string. If the string is “Jack is a boy”, it will match the „a” after the “J”. The fact that this “a” is in the middle of the word does not matter to the regex engine. If it matters to you, you will need to tell that to the regex engine by using word boundaries. We will get to that later. This regex can match the second „a” too. It will only do so when you tell the regex engine to start searching through the string after the first match. In a text editor, you can do so by using its “Find Next” or “Search Forward” function. In a programming language, there is usually a separate function that you can call to continue searching through the string after the previous match. Similarly, the regex «cat» will match „cat” in “About cats and dogs”. This regular expression consists of a series of three literal characters. This is like saying to the regex engine: find a «c», immediately followed by an «a», immediately followed by a «t». Note that regex engines are case sensitive by default. «cat» does not match “Cat”, unless you tell the regex engine to ignore differences in case.

Special Characters
Because we want to do more than simply search for literal pieces of text, we need to reserve certain characters for special use. In the regex flavors discussed in this tutorial, there are 11 characters with special meanings: the opening square bracket «[», the backslash «\», the caret «^», the dollar sign «$», the period or dot «.», the vertical bar or pipe symbol «|», the question mark «?», the asterisk or star «*», the plus sign «+», the opening round bracket «(» and the closing round bracket «)». These special characters are often called “metacharacters”. If you want to use any of these characters as a literal in a regex, you need to escape them with a backslash. If you want to match „1+1=2”, the correct regex is «1\+1=2». Otherwise, the plus sign will have a special meaning. Note that «1+1=2», with the backslash omitted, is a valid regex. So you will not get an error message. But it will not match “1+1=2”. It would match „111=2” in “123+111=234”, due to the special meaning of the plus character. If you forget to escape a special character where its use is not allowed, such as in «+1», then you will get an error message. All other characters should not be escaped with a backslash. That is because the backslash is also a special character. The backslash in combination with a literal character can create a regex token with a special meaning. E.g. «\d» will match a single digit from 0 to 9.

Special Characters and Programming Languages
If you are a programmer, you may be surprised that characters like the single quote and double quote are not special characters. That is correct. When using a regular expression or grep tool like PowerGREP or the

42 search function of a text editor like EditPad Pro, you should not escape or repeat the quote characters like you do in a programming language. In your source code, you have to keep in mind which characters get special treatment inside strings by your programming language. That is because those characters will be processed by the compiler, before the regex library sees the string. So the regex «1\+1=2» must be written as "1\\+1=2" in C++ code. The C++ compiler will turn the escaped backslash in the source code into a single backslash in the string that is passed on to the regex library. To match „c:\temp”, you need to use the regex «c:\\temp». As a string in C++ source code, this regex becomes "c:\\\\temp". Four backslashes to match a single one indeed. See the tools and languages section in this book for more information on how to use regular expressions in various programming languages.

Non-Printable Characters
You can use special character sequences to put non-printable characters in your regular expression. «\t» will match a tab character (ASCII 0x09), «\r» a carriage return (0x0D) and «\n» a line feed (0x0A). Remember that Windows text files use “\r\n” to terminate lines, while UNIX text files use “\n”. You can include any character in your regular expression if you know its hexadecimal ASCII or ANSI code for the character set that you are working with. In the Latin-1 character set, the copyright symbol is character 0xA9. So to search for the copyright symbol, you can use «\xA9». Another way to search for a tab is to use «\x09». Note that the leading zero is required.

At the 15th character in the match. The reason behind this is that the regex-directed engine is “eager”. The entire regular expression could be matched starting at character 15. Arriving at the 4th character in the match. Again. This is because certain very useful features. I will explain step by step how the regex engine actually processes that token. If the resulting match is only „regex”. «c» fails to match here and the engine carries on. All the regex flavors treated in this tutorial are based on regex-directed engines. This inside look may seem a bit long-winded at certain times. The first match is considered good enough. «t» fails to match “p”. the engine will start at the first character of the string. the engine will try to match the first token in the regex «c» to the first character in the match “H”. It will help you understand quickly why a particular regex does not do what you initially expected. lex. The result is that the regex-directed engine will return the leftmost match. and regex-directed engines. This fails too. If the result is „regex not”. It will try all possible permutations of the regular expression at the first character. you can be certain the engine is regex-directed. This fails. then it is text-directed. «c» again matches „c”. When applying «cat» to “He captured a catfish for his cat. «c» matches „c”. So it will continue with the 5th: “a”. The Regex-Directed Engine Always Returns the Leftmost Match This is a very important point to understand: a regex-directed engine will always return the leftmost match. the engine knows the regex cannot be matched starting at the 4th character in the match. flex. There are no other possible permutations of this regex. When applying a regex to a string. You can do the test by applying the regex «regex|regex not» to the string “regex not”. There are two kinds of regular expression engines: text-directed engines. it will try all possible permutations of the regex. You can easily find out whether the regex flavor you intend to use has a text-directed or regex-directed engine. But understanding how the regex engine works will enable you to use its full power and help you avoid common mistakes. At that point. Jeffrey Friedl calls them DFA and NFA engines. respectively. because it merely consists of a sequence of literal characters. Only if all possibilities have been tried and found to fail. Notable tools that use text-directed engines are awk. So the regex engine tries to match the «c» with the “e”. The engine is "eager" to report a match. MySQL and Procmail.43 3. can only be implemented in regex-directed engines. In this tutorial. egrep. It will therefore report the first three letters of catfish as a valid match. The engine never proceeds beyond this point to see if there are any “better” matches. the engine is regex-directed. But then. even if a “better” match could be found later. The engine will then try to match the second token «a» to the 5th character. This will save you lots of guesswork and head-scratching when you need to write more complex regexes. If backreferences and/or lazy quantifiers are available. . in exactly the same order. such as lazy quantifiers and backreferences. „a”. First Look at How a Regex Engine Works Internally Knowing how the regex engineworks will enable you to craft better regexes more easily. For awk and egrep. This succeeds too. will the engine continue with the second character in the text. No surprise that this kind of engine is more popular. as does matching the «c» with the space. Again.”. after introducing a new regex token. there are a few versions of these tools that use a regex-directed engine. The engine then proceeds to attempt to match the remainder of the regex at character 15 and finds that «a» matches „a” and «t» matches „t”.

Some of the results may be surprising. our regex engine simply appears to work like a regular text search routine. However. . once you know how the engine works. A text-directed engine would have returned the same result too. In following examples. the way the engine works will have a profound impact on the matches it will find.44 In this first example of the engine's internals. it is important that you can follow the steps the engine takes in your mind. But they are always logical and predetermined.

The usual metacharacters are normal characters inside a character class. «[0-9]» matches a single digit between 0 and 9. It will not match the q in the string “Iraq”. It means: “a q followed by a character that is not a u”. the backslash (\). also called “character set”. because it is the “character that is not a u” that is matched by the negated character class in the above regexp. Your regex will work fine if you escape the regular metacharacters inside a character class. It will match the q and the space after the q in “Iraq is a country”. negated character classes also match (invisible) line break characters. Useful Applications Find a word. «[0-9a-fxA-FX]» matches a hexadecimal digit or the letter X. If you want to match an a or an e. «[0-9a-fA-F]» matches a single hexadecimal digit. If you want the regex to match the q. But we will get to that later. and do not need to be escaped by a backslash. Indeed: the space will be part of the overall match. Character Classes or Character Sets With a "character class". Metacharacters Inside Character Classes Note that the only special characters or metacharacters inside a character class are the closing bracket (]). you can tell the regex engine to match only one out of several characters. Find a C-style hexadecimal number with «0[xX][A-Fa-f0-9]+». Negated Character Classes Typing a caret after the opening square bracket will negate the character class. «q[^u]» does not mean: “a q not followed by a u”. case insensitively. It is important to remember that a negated character class still must match a character. «gr[ae]y» will not match “graay”. The order of the characters inside a character class does not matter. “graey” or any such thing. such as «sep[ae]r[ae]te» or «li[cs]en[cs]e». A character class matches only a single character. Unlike the dot. Simply place the characters you want to match between square brackets. and only the q. use «[ae]». You can use more than one range. the caret (^) and the hyphen (-). Again. even if it is misspelled. You can combine ranges and single characters. Find an identifier in a programming language with «[A-Za-z_][A-Za-z_0-9]*». use «[+*]». You can use a hyphen inside a character class to specify a range of characters. the order of the characters and the ranges does not matter. The results are identical.45 4. You could use this in «gr[ae]y» to match either „gray” or „grey”. in both strings. but doing so significantly reduces readability. you need to use negative lookahead: «q(?!u)». . To search for a star or plus. The result is that the character class will match any character that is not in the character class. Very useful if you do not know whether the document you are searching through is written in American or British English.

«[\da-fA-F]» matches a hexadecimal digit. When applied to “1 + 2 = 3”. «[\s\d]» matches a single character that is either whitespace or a digit. a series of shorthand character classes are available. In most flavors. In EditPad Pro. In all flavors. it includes «[ \t]». characters with diacritics used in languages such as French and Spanish will be included. If you are using the Western script.46 To include a backslash as a character without any special meaning inside a character class. Both «[-x]» and «[x-]» match an x or a hyphen. Negated Shorthand Character Classes The above three shorthands also have negated versions. In most. the caret (^) and the hyphen (-) can be included by escaping them with a backslash. word characters from other languages may also match. Shorthand character classes can be used both inside and outside the square brackets. while the latter matches „1” (one). In some flavors. it will include «[A-Za-z]». «[\\x]» matches a backslash or an x. since it improves readability. «[x^]» matches an x or a caret. I recommend the latter method. «\s\d» matches a whitespace character followed by a digit. That is: «\s» will match a space or a tab. Again. you can see the characters matched by «\w» in PowerGREP when using the Western script. Exactly which characters it matches differs between regex flavors. for example. The best way to find out is to do a couple of tests with the regex flavor you are using. or by placing them in a position where they do not take on their special meaning. «\D» is the same as «[^\d]». etc. Shorthand Character Classes Since certain character classes are used often. it also includes a carriage return or a line feed as in «[ \t\r\n]». You can put the closing bracket right after the opening bracket. «\s» stands for “whitespace character”. and is equivalent to «[0-9a-fA-F]». Russian characters will be included. the former regex will match „ 2” (space two). you have to escape it with another backslash. which characters this actually includes. depends on the regex flavor. In all flavors discussed in this tutorial. or right before the closing bracket. The hyphen can be included right after the opening bracket. the underscore and digits are also included. or the negating caret. «\w» stands for “word character”. To include a caret. the actual character range depends on the script you have chosen in Options|Font. rarely used non-printable characters such as vertical tab and form feed. «\W» is short for «[^\w]» and «\S» is the equivalent of «[^\s]». The closing bracket (]). «\d» is short for «[0-9]». Some flavors include additional. If you are using the Cyrillic script. In the screen shot. «[^]x]» matches any character that is not a closing bracket. . or right after the negating caret. place it anywhere except right after the opening bracket. «[]x]» matches a closing bracket or an x.

But because we are using a regex-directed engine. digit. because that is the leftmost match. If you want to repeat the matched character. which matches the next character in the text. however. it will match „3333” in the middle of this string. or is not whitespace. you need to use lookahead and lookbehind. Repeating Character Classes If you repeat a character class by using the «?». Let us take a look at that first. The engine will then try to match the remainder of the regex with the text. The regex «[0-9]+» can match „837” as well as „222”. and continue with the next character in the string. whitespace or otherwise. The next token in the regex is the literal «r». and not just the character that it matched. the leftmost match was returned. «[\D\S]» will match any character. which can be matched with the following character as well. But the engine simply did not get that far. Again. When applied to the string “833337”. But I digress. Looking Inside The Regex Engine As I already said: the order of the characters inside a character class does not matter. Nothing noteworthy happens for the first twelve characters in the string. The engine has found a complete match with the text starting at character 13. you will need to use backreferences. When the engine arrives at the 13th character. but not “8”. and „gray” could have been matched in the string. «([09])\1+» will match „222” but not “837”. The last regex token is «y». It will return „grey” as the match result. even though we put the «a» first in the character class. it must continue trying to match all the other permutations of the regex pattern before deciding that the regex cannot be matched with the text starting at character 13. «[ae]» is attempted at the next character in the text (“e”). and fail. The latter will match any character that is not a digit or whitespace. That is: «gr[ae]y» can match both „gray” and „grey”. The character class gives the engine two options: match «a» or match «e». It will first attempt to match «a». and whitespace is not a digit. So it will match „x”. you will repeat the entire character class. The engine will fail to match «g» at every step. and find that «e» matches „e”. rather than the class. «gr[ae]y» will match „grey” in “Is his hair grey or gray?”. and look no further. . Because a digit is not whitespace. Below. „g” is matched. So the third token. «*» or «+» operators.47 Be careful when using the negated shorthands inside square brackets. «[\D\S]» is not the same as «[^\d\s]». We already saw how the engine applies a regex consisting only of literal characters. The former. If you do not want that. will match any character that is either not a digit. because another equally valid match was found to the left of it. I did not yet explain how character classes work inside the regex engine. I will explain how it applies a regex that has more than one permutation. So it will continue with the other option.

]\d\d» is a better solution. Other languages and regex libraries have adopted Perl's terminology. and single-line mode only affects the dot. In this match. the first dot matched „5”. «\d\d[. Let's say we want to match a date in mm/dd/yy format. In all regex flavors discussed in this tutorial. the string could never contain newlines. it is also the most commonly misused metacharacter. Seems fine at first. Multi-line mode only affects anchors. This exception exists mostly because of historic reasons. . and the second matched „7”. like this: m/^regex$/s. In all programming languages and regex libraries I know. without caring what that character is.Singleline).Singleline. and everything will match just fine when you test the regex on valid data. This regex allows a dash. This is a bit unfortunate. The problem is that the regex will also match in cases where it should not match. but we want to leave the user the choice of date separators.\d\d».48 5. In Perl. Put in a dot. please give it a clearer label like was done in RegexBuddy. You can activate single-line mode by adding an s after the regex code. The only exception are newlinecharacters. It allows you to be lazy. so we do not need to escape it with a backslash./. The effect is that with these tools. They would read a file line by line. and apply the regular expression separately to each line. space.]\d\d[. you simply tick the checkbox labeled “dot matches newline”. Modern tools and languages can apply regular expressions to very large strings or even entire files./. The dot matches a single character. It will match a date like „02/12/03” just fine. including newlines. because it is easy to mix up this term with “multi-line mode”. So if you expose this option to your users. the dot or period is one of the most commonly used metacharacters. Unfortunately. Obviously not what we intended. In RegexBuddy. the dot is short for the negated character class «[^\n]» (UNIX regex flavors) or «[^\r\n]» (Windows regex flavors). EditPad Pro and PowerGREP. so the dot could never match them. All regex flavors discussed here have an option to make the dot match all characters. The first tools that used regular expressions were line-based. such as in Regex. Trouble is: „02512703” is also considered a valid date by this regular expression. RegexOptions.\d\d. EditPad Pro or PowerGREP. the dot will not match a newline character by default. I will illustrate this with a simple example. So by default.Match(“string”. Use The Dot Sparingly The dot is a very powerful regex metacharacter. activating single-line mode has no effect other than making the dot match newlines. the mode where the dot also matches newlines is called "single-line mode". you activate this mode by specifying RegexOptions. Remember that the dot is not a metacharacter inside a character class. If you are new to regular expressions. dot and forward slash as date separators. “regex”. The Dot Matches (Almost) Any Character In regular expressions. some of these cases may not be so obvious at first. The quick solution is «\d\d.. When using the regex classes of the .NET framework.

If you are validating user input. The regex matches „“string one” and “string two””. How perfect you want your regex to be depends on what you want to do with it./. Now go ahead and test it on “Houston.][0-3]\d[/. We want any number of characters that are not double quotes or newlines between the quotes. we improved our regex by replacing the dot with a character class. but the warning is important enough to mention it here as well. we have a problem with “string one” and “string two”. Suppose you want to match a double-quoted string.” Ouch. our last attempt is probably more than sufficient to parse the data without errors. Definitely not what we intended. it has to be perfect. we will do the same. Use Negated Character Sets Instead of the Dot I will explain this in depth when I present you the repeat operators star and plus.49 This regex is still far from perfect.]\d\d» is a step ahead. and the star allows the dot to be repeated any number of times. Please respond. We can have any number of any character between the double quotes.*"» seems to do the trick just fine. So the proper regex is «"[^"\r\n]*"». Our original definition of a double-quoted string was faulty. The reason for this is that the star is greedy. including zero. If you test this regex on “Put a “string” between double quotes”. so «". «[0-1]\d[. . The dot matches any character. If you are parsing data files from a known source that generates its files in the same way every time. You can find a better regex to match dates in the example section. Sounds easy. it will match „“string”” just fine. It matches „99/99/99” as a valid date. I will illustrate with an example. In the date-matching example. We do not want any number of any character between the quotes. though it will still match „19/39/99”. Here.

such as in Regex. after or between characters. It is traditionally called "multi-line mode". and also before every line break (between “e” and “\n”). «^b» will not match “abc” at all. See below for the inside view of the regex engine. «$» matches right after the last character in the string. «^\s+» matches leading whitespace and «\s+$» matches trailing whitespace. using anchors is very important.Multiline. Similarly. Instead.50 6. matched by «^». “regex”. I have explained literal characters and character classes. because «\d+» matches the 4.NET. They can be used to “anchor” the regex match at a certain position. . and “end of string” must be matched right after it. RegexOptions. Anchors are a different breed. putting one in a regex will cause the regex engine to try to match a single character. They do not match any character at all. «c$» matches „c” in “abc”. you have to explicitly activate this extended functionality. So before validating input. It is easy for the user to accidentally type in a space. The correct regex to use is «^\d+$». Useful Applications When using regular expressions in a programming language to validate user input. it is good practice to trim leading and trailing whitespace. and regex tools like PowerGREP. rather than short strings. When Perl reads from a line from a text file. Using ^ and $ as Start of Line and End of Line Anchors If you have a string consisting of multiple lines. it will accept the input even if the user entered “qsdf4ghjk”. the anchors match before and after newlines when you specify RegexOptions.Match(“string”. while «a$» does not match at all. the caret and dollar always match at the start and end of each line. the line break will also be stored in the variable. because the «b» cannot be matched right after the start of the string. This makes sense because those applications are designed to work with entire files. «^» can then match at the start of the string (before the “f” in the above string). like this: m/^regex$/m. you do this by adding an m after the regex code. Applying «^a» to “abc” matches „a”. as well as after each line break (between “\n” and “s”). The caret «^» matches the position before the first character in the string. In . they match a position before. Therefore. it is often desirable to work with lines.. In every programming language and regex library I know. «$» will still match at the end of the string (after the last “e”). the entire string must consist of digits for «^\d+$» to be able to match. Handy use of alternation and /g allows us to do this in a single line of code. Start of String and End of String Anchors Thus far. Likewise. all the regex engines discussed in this tutorial have the option to expand the meaning of both anchors. you could use $input =~ s/^\s+|\s+$//g.Multiline). In Perl. rather than the entire string. Because “start of string” must be matched before the match of «\d+». In Perl. In both cases. If you use the code if ($input =~ m/\d+/) in a Perl script to see if the user entered an integer number. In text editors like EditPad Pro or GNU Emacs. like “first line\nsecond line” (where \n indicates a line break).

«\z» matches after the line break. As usual. In VB. both «^[a-z]+$» and «\A[a-z]+\Z» will match „joe”. including Java. When applied to this string. Looking Inside the Regex Engine Let's see what happens when we try to match «^4$» to “749\n486\n4” (where \n represents a newline character) in multi-line mode. the regex engine starts at the first character: “7”. See below. However. RegexOptions. There are no other permutations of the . These two tokens never match at line breaks. just like we want it. Depending on the situation. Reading a line from a file with the text “joe” results in the string “joe\n”. and is copied by many regex flavors. the resulting string will end with a line break. nothing is deleted. rather than matching a character. it can result in a zero-length match. However. the match does include a starting position. The engine then advances to the next regex token: «4». Likewise.NET and PCRE.Replace method will remove the regex match from the string. In Perl. If you only want a match at the absolute very end of the string. "^". . It remains at “7”. matching only a position can be very useful. «\A» and «\Z» only match at the start and the end of the entire file. If the string ends with a line break. This “enhancement” was introduced by Perl. even when you turn on “multiline mode”. Using «^\d*$» to test if the user entered a number (notice the use of the star instead of the plus). Zero-Length Matches We saw that the anchors match at a position. Since this token is a zero-width token.Replace(Original. the engine does not try to match it with the character.51 Permanent Start of String and End of String Anchors «\A» only ever matches at the start of the string. rather than at the very end of the string.NET. In email. Since the match does not include any characters. there is one exception. it is common to prepend a “greater than” symbol and a space to each line of the quoted message. would cause the script to accept an empty string as a valid input. when reading a line from a file. so the regex «^» matches at the start of the quoted message. where the caret and dollar always match at the start and end of lines. «\A[a-z]+\z» does not match “joe\n”. This means that when a regex only consists of one or more anchors. use «\z» (lower case z instead of upper case Z). for example. which is not matched by the character class. then «\Z» and «$» will match at the position before that line break. In EditPad Pro and PowerGREP. and insert the replacement string (greater than symbol and a space). which does not match “7”. The Regex. this can be very useful or undesirable. The first token in the regular expression is «^». We are using multi-line mode. Since the previous token was zero-width. "> ". we can easily do this with Dim Quoted as String = Regex. «4» is a literal character. «\Z» only ever matches at the end of the string. and the replacement string is inserted there. This is true in all regex flavors discussed in this tutorial. the regex engine does not advance to the next character in the string. Strings Ending with a Line Break Even though «\Z» and «$» only match at the end of the string (when the option for the caret and dollar to match at embedded line breaks is off).Multiline). «^» indeed matches the position before “7”. and after each newline. but rather with the position before the character that the regex engine has reached so far.

It must be either a newline. and the mighty dollar is a strange beast. so the engine continues at the next character. We already saw that those match. the position before “\n” is preceded by a character. optional. . Again. at the next character: “4”. Another Inside Look Earlier I mentioned that «^\d*$» would successfully match an empty string. This time. There is only one “character” position in an empty string: the void after the string. and that character is not a newline. and that character is not a newline. The next attempt. the dollar matches successfully. After that. It matches the position before the void after the string. Not even a negated character class. and fails again. or the length+1 if string indices are one-based in your programming language. but does not advance the character position in the string. it would return the length of the string if string indices are zero-based. one of the star's effects is that it makes the «\d». As we will see later. but the star turns the failure of the «\d» into a zero-width success. Same at the six and the newline. Since that is the case after the example. because it is preceded by the void before the string. “8”. so the engine starts again with the first regex token. The engine continues at “9”. What you have to watch out for is that String[Regex.52 regex. Finally. Let's see why. «4» matches „4”. where the caret does not match. So the engine arrives at «$». The engine will proceed with the next regex token. The next token is «\d*». Then. That fails. the engine has found a successful match: the last „4” in the string. Yet again. With success. the dollar will check the current character. In fact. If you would query the engine for the character position. the regex engine tries to match the first token at the third “4” in the string. The current regex token is advanced to «$». the entire regex has matched the empty string. the regex engine arrives at the second “4” in the string. “9”. This position is preceded by a character. No regex token that needs a character to match can match here. It is zero-width. without advancing the position in the string. or the void after the string. it would return zero. At this point. and the void after the string. because this position is preceded by a character. and the current character is advanced to the very last position in the string: the void after the string. However. at “\n”. and that character is not a newline. it was successfully matched at the second “4”. Again. This can also happen with «^» and «^$» if the last character in the string is a newline. because MatchPosition can point to the void after the string. and the engine reports success. The dollar cannot match here. Since «$» was the last token in the regex. also fails. because it is preceded by a newline character. Previously. so it will try to match the position before the current character. Now the engine attempts to match «$» at the position before (indeed: before) the “8”. the regex engine advances to the next regex token. the engine successfully matches «4» with „4”. «4». The «^» can match at the position before the “4”. If you would query the engine for the length of the match. The engine will try to match «\d» with the void after the string. for «$» to match the position before the current character. It does not matter that this “character” is the void after the string. in this case. «^» cannot match at the position before the 4. we are trying to match a dollar sign.MatchPosition] may cause an access violation or segmentation fault. The first token in the regex is «^». Caution for Programmers A regular expression such as «$» all by itself can indeed match after the string. and the engine advances both the regex token and the string character. the engine must try to match the first token again.

because the T is a word character and the character before it is the void before the start of the string. Between a non-word character and a word character following right after the non-word character. This is because any position between characters can never be both at the start and at the end of a word. «i» does not match “T”. there is only one metacharacter that matches both before a word and after a word. so the engine retries the first token at the next character position. Looking Inside the Regex Engine Let's see what happens when we apply the regex «\bis\b» to the string “This island is beautiful”. The engine continues with the next token: the literal «i». Simply put: «\b» allows you to perform a “whole words only” search using a regular expression in the form of «\bword\b». . and neither between the “i” and the “s”. All non-word characters are always matched by «\W». Between a word character and a non-word character following right after the word character. if the first character is a word character. Using only one operator makes things easier for you. because the previous regex token was zero-width. and the preceding character is. Again. «\B» matches at any position between two word characters as well as at any position between two non-word characters. There are four different positions that qualify as word boundaries: • • • • Before the first character in the string. if the last character is a word character. Note that «\w» usually also matches digits. The next character in the string is a space. «\B» matches at every position where «\b» does not.53 7. So saying "«\b» matches before and after an alphanumeric sequence“ is more exact than saying ”before and after a word". but all word characters are always matched by the short-hand character class «\w». This regex will not match “44 sheets of a4”. In Perl and the other regex flavors discussed in this tutorial. «\b» matches here because the space is not a word character. It matches at a position that is called a “word boundary”. A “word character” is a character that can be used to form words. All characters that are not “word characters” are “non-word characters”. This match is zero-length. «\b» matches here. The exact list of characters is different for each regex flavor. So «\b4\b» can be used to match a 4 that is not part of a larger number. Effectively. Word Boundaries The metacharacter «\b» is an anchor like the caret and the dollar sign. The engine does not advance to the next character in the string. The engine starts with the first token «\b» at the first character “T”. «\b» cannot match at the position between the “T” and the “h”. Negated Word Boundary «\B» is the negated version of «\b». After the last character in the string. the position before the character is inspected. It cannot match between the “h” and the “i” either. the engine continues with the «i» which does not match with the space. Since this token is zero-length.

The engine has successfully matched the word „is” in our string. the engine tries to match the second «\b» at the position before the “l”. also matches at the position before the second space in the string because the space is not a word character. Again. . But «\b» matches at the position before the third “i” in the string. Continuing. «\b» matches between the space and the second “i” in the string. The engine reverts to the start of the regex and advances one character to the “s” in “island”. «\b». skipping the two earlier occurrences of the characters i and s. and finds that «i» matches „i” and «s» matches «s». it would have matched the „is” in “This”. The engine continues. and the character before it is. Now. the regex engine finds that «i» matches „i” and «s» matches „s”. the «\b» fails to match and continues to do so until the second space is reached.54 Advancing a character and restarting with the first regex token. It matches there. If we had used the regular expression «is». The last token in the regex. but matching the «i» fails. This fails because this position is between two word characters.

we would need to use «\b(cat|dog)\b». or. Suppose you want to use a regex to match a list of function names in a programming language: Get. or everything to the right of the vertical bar. If we use «GetValue|Get|SetValue|Set». Remember That The Regex Engine Is Eager I already explained that the regex engine is eager. We do not want to match Set or SetValue if the string is “SetValueFunction”. and the engine continues with the next character in the string. If you want to search for the literal text «cat» or «dog». «SetValue» will be attempted before «Set». and at the first character in the string. then either “cat” or “dog”. So it continues with the second option. there are no other tokens in the regex outside the alternation. “S”. «G». and change the order of the options. At this point. The best option is probably to express the fact that we only want to match complete words. So the solution is «\b(Get|GetValue| Set|SetValue)\b» or «\b(Get(Value)?|Set(Value)?)\b». In this example. we can optimize this further to «\b(Get|Set)(Value)?\b». Since all options have the same end. it tells the regex engine to match either everything to the left of the vertical bar. Set or SetValue. as well as the next token in the regex. The match fails again. the order of the alternatives matters. There are several solutions. Because the regex engine is eager. Because the question mark is greedy. «t» matches „t”. Alternation with The Vertical Bar or Pipe Symbol I already explained how you can use character classes to match a single character out of several possible characters. The match fails. The consequence is that in certain situations. being the second «G» in the regex. simply expand the list: «cat|dog|mouse|fish». The next token is the first «S» in the regex. it considers the entire alternation to have been successfully matched as soon as one of the options has. So it knows that this regular expression uses alternation. . and the engine will match the entire string. That is. you will need to use round brackets for grouping. The alternation operator has the lowest precedence of all regex operators. «SetValue» will be attempted before «Set». This tells the regex engine to find a word boundary. One option is to take into account that the regex engine is eager. The next token in the regex is the «e» after the «S» that just successfully matched. the third option in the alternation has been successfully matched. the regex engine would have searched for “a word boundary followed by cat”. and that the entire regex has not failed yet. We could also combine the four options into two and use the question mark to make part of them optional: «Get(Value)?|Set(Value)?». so the entire regex has successfully matched „Set” in “SetValue”.55 8. the regex did not match the entire string. the regex engine studied the entire regular expression before starting. If we want to improve the first example to match whole words only. If you want more options. «e» matches „e”. separate both options with a vertical bar or pipe symbol: «cat|dog». You can use alternation to match a single regular expression out of several possible regular expressions. "dog followed by a word boundary. It will stop searching as soon as it finds a valid match. The match succeeds. The regex engine starts at the first token in the regex. If you want to limit the reach of the alternation. However. and then another word boundary. The next token. Alternation is similar. Let's see how this works out when the string is “SetValue”. Contrary to what we intended. If we had omitted the round brackets. The obvious solution is «Get|GetValue|Set|SetValue». GetValue.

The first token in the regex is the literal «c». However. „February 23”. and finds that «o» matches „o”. Now the engine checks whether «u» matches “r”. will the engine try ignoring the part the question mark applies to. «Feb(ruary)? 23(rd)?» matches „February 23rd”. or do not try to match it. . and placing the question mark after the closing bracket. The question mark allows the engine to continue with «r». This matches „r” and the engine reports that the regex successfully matched „color” in our string. E. You can make the question mark lazy (i. The engine will always try to match that part. turn off the greediness) by putting a second question mark after the first.g. But this fails to match “n” as well. and «o». E. Only if this causes the entire regular expression to fail. Now. You can make several tokens optional by grouping them together using round brackets.g. the match will always be „Feb 23rd” and not „Feb 23”.: «Nov(ember)?» will match „Nov” and „November”. Again: no problem. You can write a regular expression that matches many alternatives by including more than one question mark. Then the engine checks whether «u» matches “n”. „Feb 23rd” and „Feb 23”. I have introduced the first metacharacter that is greedy. The question mark gives the regex engine two choices: try to match the part the question mark applies to. The effect is that if you apply the regex «Feb 23(rd)?» to the string “Today is Feb 23rd.56 9. I will say a lot more about greediness when discussing the other repetition operators. Therefore. the question mark tells the regex engine that failing to match «u» is acceptable. This fails.e. After a series of failures. The first position where it matches successfully is the „c” in “colonel”. the engine starts again trying to match «c» to the first o in “colonel”. the engine can only conclude that the entire regular expression cannot be matched starting at the „c” in “colonel”. Optional Items The question mark makes the preceding token in the regular expression optional.: «colou?r» matches both „colour” and „color”. The engine continues. Therefore. This fails. 2003”. «l» and «o» match the following characters. Looking Inside The Regex Engine Let's apply the regular expression «colou?r» to the string “The colonel likes the color green”. «l» matches „l” and another «o» matches „o”. the engine will skip ahead to the next regex token: «r». «c» will match with the „c” in “color”. Important Regex Concept: Greediness With the question mark.

I did not. so the regular expression does not need to exclude any invalid use of sharp brackets. and proceed with the remainder of the regex. You could use «\b[1-9][0-9]{3}\b» to match a number between 1000 and 9999. The second character class matches a letter or digit. If it sits between sharp brackets. Because we used the star. The star repeats the second character class. Most people new to regular expressions will attempt to use «<.}» is the same as «+». «<[A-Za-z][A-Za-z0-9]*>» matches an HTML tag without any attributes.57 10. it's OK if the second character class matches nothing. it is an HTML tag. The sharp brackets are literals. the maximum number of matches is infinite. You might expect the regex to match „<EM>” and when continuing after that match. it will go back to the plus. like those discussed in this tutorial. The syntax is {min. Limiting Repetition Modern regex flavors. The reason is that the plus is greedy. Only if that causes the entire regex to fail. I could also have used «<[A-Za-z0-9]+>». So «{0. Obviously not what we wanted. «\b[1-9][09]{2. . which is not a valid HTML tag. So our regex will match a tag like „<B>”.max}. where min is a positive integer number indicating the minimum number of matches. They will be surprised when they test it on a string like “This is a <EM>first</EM> test”. the star and the repetition using curly braces are greedy. Notice the use of the word boundaries. That is. If the comma is present but max is omitted.+>». The star will cause the second character class to be repeated three times.}» is the same as «*». That is. Repetition with Star and Plus I already introduced one repetition operator or quantifier: the question mark. The first character class matches a letter. When matching „<HTML>”. Omitting both the comma and max tells the engine to repeat the token exactly min times. Watch Out for The Greediness! Suppose you want to use a regex to match an HTML tag. After that. matching „T”. The regex will match „<EM>first</EM>”. The plus tells the engine to attempt to match the preceding token once or more.4}\b» matches a number between 100 and 99999. and «{1. But it does not. will the regex engine backtrack. the plus causes the regex engine to repeat the preceding token as often as possible. Let's take a look inside the regex engine to see in detail how this works and why this causes our regex to fail. You know that the input will be a valid HTML file. It tells the engine to attempt match the preceding token zero times or once. have an additional repetition operator that allows you to specify how many times a token can be repeated. because this regex would match „<1>”. But this regex may be sufficient if you know the string you are searching through does not contain any such invalid tags. I will present you with two possible solutions. the first character class will match „H”. Like the plus. and max is an integer equal to or greater than min indicating the maximum number of matches. make it give up the last iteration. „</EM>”. The asterisk or star tells the engine to attempt to match the preceding token zero or more times. in effect making it optional. „M” and „L” with each step.

The total match so far is reduced to „<EM>first</EM> te”. Let's have another look inside the regex engine. the engine has to backtrack for each character in the HTML tag that it is trying to match. Remember that the regex engine is eager to return a match.+» is reduced to „EM>first</EM> tes”. Again. Laziness Instead of Greediness The quick fix to this problem is to make the plus lazy instead of greedy. there is a better option than making the plus lazy. «>» can match the next character in the string. this time repeated by a lazy plus. the engine will backtrack. Again. So our example becomes «<. and the engine continues with «>» and “M”.+» has matched „<EM>first</EM> test” and the engine has arrived at the end of the string. «<. It will reduce the repetition of the plus by one. This tells the regex engine to repeat the dot as few times as possible. and the engine continues repeating the dot. The next token is the dot. which matches any character except newlines. Therefore. The plus is greedy. It will report the first valid match it finds. We can use a greedy plus and a negated character class: «<[^>]+>». The engine reports that „<EM>” has been successfully matched. Now. So the engine continues backtracking until the match of «. The next token in the regex is still «>». This is a literal. When using the negated character class. The last token in the regex has been matched. The engine remembers that the plus has repeated the dot more often than is required. Only at this point does the regex engine continue with the next token: «>». these cannot match. Because of greediness. As we already know. the first place where it will match is the first „<” in the string. the engine will repeat the dot as many times as it can. So the match of «. The reason why this is better is because of the backtracking.) Rather than admitting failure. The engine reports that „<EM>first</EM>” has been successfully matched. no backtracking occurs at all when the string contains valid HTML code. „M” is matched. So the match of «. the curly braces and the question mark itself. Again. That's more like it. (Remember that the plus requires the dot to match only once. and the dot is repeated once more. Now. The dot is repeated by the plus. You should see the problem by now.+» is expanded to „EM”. But «>» still cannot match. When using the lazy plus. The next token is the dot.58 Looking Inside The Regex Engine The first token in the regex is «<».+?>». The minimum is one. You can do the same with the star. The next character is the “>”. «>» cannot match here. . causing the engine to backtrack further. The dot will match all remaining characters in the string. So the engine matches the dot with „E”. so the regex continues to try to match the dot with the next character. «<» matches the first „<” in the string. this is the leftmost longest match. It will not continue backtracking further to see if there is another possible match. and then continue trying the remainder of the regex. „>” is matched successfully. An Alternative to Laziness In this case.+» is reduced to „EM>first</EM”. This fails. The requirement has been met. But now the next character in the string is the last “t”. The dot fails when the engine has reached the void after the end of the string. The last token in the regex has been matched. the engine will backtrack. But this time. So far. The dot matches the „>”. the backtracking will force the lazy plus to expand rather than reduce its reach. The dot matches „E”. You can do that by putting a question markbehind the plus in the regex. and the engine tries again to continue with «>».

They do not get the speed penalty. Finally. or perhaps in a custom syntax coloring scheme for EditPad Pro. You will not notice the difference when doing a single search in a text editor. remember that this tutorial only talks about regex-directed engines. But you will save plenty of CPU cycles when using such a regex is used repeatedly in a tight loop in a script that you are writing. but they also do not support lazy repetition operators. Text-directed engines do not backtrack. .59 Backtracking slows down the regex engine.

Note that only round brackets can be used for grouping. Use Round Brackets for Grouping By placing part of a regular expression inside round brackets or parentheses. \U1 inserts the first backreference in uppercase. at the expense of making your regular expression slightly harder to read. the first backreference will contain „Value”. In the first case. you can use the backreference in the replacement text during a search-and-replace operation by typing \1 (backslash one) into the replacement text. EditPad Pro and PowerGREP have a unique feature that allows you to change the case of the backreference. the actual replacement will be “Lite version” in case „EditPad Lite” was matched. because it did not match anything. \L1 in lowercase and \F1 with the first character in uppercase and the remainder in lowercase. How to Use Backreferences Backreferences allow you to reuse part of the regex match. and “Pro version” in case „EditPad Pro” was matched. unless you use non-capturing parentheses. You can reuse it inside the regular expression (see below).60 11. Remembering part of the regex match in a backreference. slows down the regex engine because it has more work to do. Square brackets define a character class. That question mark is the regex operator that makes the previous token optional. e. you can speed things up by using non-capturing parentheses. If you searched for «EditPad (Lite|Pro)» and use “\1 version” as the replacement. or afterwards. round brackets also create a “backreference”. The regex «Set(Value)?» matches „Set” or „SetValue”. In EditPad Pro or PowerGREP. and curly braces are used by a special repetition operator.g. If you do not use the backreference. I have already used round brackets for this purpose in previous topics throughout this tutorial. A backreference stores the part of the string matched by the part of the regular expression inside the parentheses. Finally. you can optimize this regular expression into «Set(?:Value)?». the first backreference will be empty. and the other letters in lowercase. Round Brackets Create a Backreference Besides grouping part of a regular expression together. to the entire group. That is. Therefore. Note the question mark after the opening bracket is unrelated to the question mark at the end of the regex. What you can do with it afterwards. you can group that part of the regular expression together. there is no confusion between the question mark as an operator to make a token optional. \I1 inserts it with the first letter of each word capitalized. depends on the tool you are using. This allows you to apply a regex operator. a repetition operator. This operator cannot appear after an opening round bracket. If you do not use the backreference. The colon indicates that the change we want to make is to turn off capturing the backreference. because an opening bracket by itself is not a valid regex token. and the question mark as a character to change the properties of a pair of round brackets. . The question mark and the colon after the opening round bracket are the special syntax that you can use to tell the regex engine that this pair of brackets should not create a backreference. In the second case.

By putting the opening tag into a backreference.61 Regex libraries in programming languages also provide access to the backreference. and the text in between. the item with index zero holds the entire regex match. we can reuse the name of the tag for the closing tag. \0 cannot be used inside a regex. etc. Libraries like . To get the string matched by the third backreference in C#. «([a-c])x\1x\1» will match „axaxa”. The «/» before it is simply the forward slash in the closing HTML tag that we are trying to match. $2.*?</\1>».Value. In . Therefore. you can use the Match object that is returned by the Match method of the Regex class. $2. The first bracket starts backreference number one. Suppose you want to match a pair of opening and closing HTML tags. You can reuse the same backreference more than once. the magic variable $& holds the entire regex match. . In Perl. which is a collection of Group objects. you can use MyMatch. If a backreference was not used in a particular match attempt (such as in the first example where the question mark made the first backreference optional). The Entire Regex Match As Backreference Zero Certain tools make the entire regex match available as backreference zero. to access the part of the string matched by the backreference.NET (dot net) Regex class also has a method Replace that can do a regex-based search-and-replace on a string. A backreference cannot be used inside itself. Depending on your regex flavor. This backreference is reused with «\1» (backslash one). the second number two. etc. In Perl. Here's how: «<([A-Z][A-Z09]*)[^>]*>. The . you can use $1. it will either give an error message. you can use the magic variables $1.NET (dot net) where backreferences are made available as an array or numbered list. «([abc]\1)» will not work.Groups[3]. To figure out the number of a particular backreference. This regex contains only one pair of parentheses. In the replacement text. or it will fail to match anything without an error message. which capture the string matched by «[A-Z][A-Z0-9]» into the first backreference. scan the regular expression from left to right and count the opening round brackets. This can be very useful when modifying a complex regular expression. Using an empty backreference in the regex is perfectly fine. because that would force the engine to continuously keep an extra copy of the entire regex match.NET (dot net). This fact means that non-capturing parentheses have another benefit: you can insert them into a regular expression without changing the numbers assigned to the backreferences. This object has a property called Groups. it is simply empty. you can use the entire regex match in the replacement text during a search and replace operation by typing \0 (backslash zero) into the replacement text. Non-capturing parentheses are not counted. In EditPad Pro or PowerGREP. Using Backreferences in The Regular Expression Backreferences can not only be used after a match has been found. to insert backreferences. Using backreference zero is more efficient than putting an extra pair of round brackets around the entire regex. but also during the match. It will simply be replaced with nothingness. etc. „bxbxb” and „cxcxc”. only in the replacement.

This match fails. The second time „a” and the third time „b”. The engine arrives again at «\1». and the dot consumes the third “<” in the string. this is not a problem. the first regex will put „cab” into the first backreference. «B» matches „B”. so the engine again backtracks. «<» matches the third „<” in the string. «[A-Z]» matches „B”. After storing the backreference. Each time. the previously saved match is overwritten. The dot matches the second „<” in the string. because of another star. and the engine is forced to backtrack to the dot. It will use the last match saved into the backreference each time it needs to be used. The star is still lazy. because of the star. so „B” it is. the regex engine will initially skip this token. This also means that «([abc]+)=\1» will match „cab=cab”. This does not match “I”. Again. Though both successfully match „cab”. Every time the engine arrives at the backreference. it will read the value that was stored. The next token is «[A-Z]». so „b” remains. The next token is «\1». and not «B». This step crosses the closing bracket of the first pair of capturing parentheses. The next token is a dot. the new value stored in the first backreference would be used. But this did not happen here. taking note that it should backtrack in case the remainder of the regex fails. and that «([abc])+=\1» will not. There is a clear difference between «([abc]+)» and «([abc])+». This fails to match at “I”.62 Looking Inside The Regex Engine Let's see how the regex engine applies the above regex to the string “Testing <B><I>bold italic</I></B> text”. «[^>]» does not match „>”. the previous value was overwritten. so the engine again takes note of the available backtracking position and advances to «<» and “I”. The position in the string remains at “>”. If a new match is found by capturing parentheses. These do not match. A complete match has been found: „<B><I>bold italic</I></B>”. However. Because of the laziness. that's perfectly fine. the regex engine does not permanently substitute backreferences in the regular expression. and position in the regex is advanced to «>». „B” is stored. The regex engine also takes note that it is now inside the first pair of capturing parentheses. Obvious when you look at a . The regex engine will traverse the string until it can match at the first „<” in the string. In this case. The engine advances to «[A-Z0-9]» and “>”. The backtracking continues until the dot has consumed „<I>bold italic”. the engine proceeds with the match attempt. The reason is that when the engine arrives at «\1». This means that if the engine had backtracked beyond the first pair of capturing parentheses before arriving the second time at «\1». «>» matches „>”. Repetition and Backreferences As I mentioned in the above inside look. and the second “<” in the string. These match. it holds «b» which fails to match “c”. The position in the string remains at “>”. At this point. The last token in the regex. This prompts the regex engine to store what was matched inside them into the first backreference. The first time. At this point. The position in the regex is advanced to «[^>]». The engine has now arrived at the second «<» in the regex. „c” was stored. while the second regex will only store „b”. Note that the token the backreference. repeated by a lazy star. The backreference still holds „B”. The first token in the regex is the literal «<». That is because in the second regex. so the engine backtracks again. Backtracking continues again until the dot has consumed „<I>bold italic</I>”. «<» matches „<” and «/» matches „/”. the plus caused the pair of parentheses to repeat three times. These obviously match. The engine does not substitute the backreference in the regular expression. and the next token is «/» which matches “/”. The next token is «/».

„b”. doubled words such as “the the” easily creep in. always double check that you are really capturing what you want. Parentheses and Backreferences Cannot Be Used Inside Character Classes Round brackets cannot be used inside character classes. „(” and „)”. So this regex will match an a followed by either «\x01» or a «b». . it is treated as a literal character. When you put a round bracket in a character class. you can easily find them. To delete the second word. but a common cause of difficulty with regular expressions nonetheless. simply type in “\1” as the replacement text and click the Replace button. When using backreferences. So the regex «[(a)b]» matches „a”. Using the regex «\b(\w+)\s+\1\b» in your text editor. Useful Example: Checking for Doubled Words When editing text. Backreferences also cannot be used inside a character class.63 simple example like this one. at least not as metacharacters. The \1 in regex like «(a)[\1b]» will be interpreted as an octal escape in most regex flavors.

starting with one.NET offers two syntaxes to create a capturing group: one using sharp brackets. the numbering can get a little confusing. or to use part of the regex match for further processing. «(?P<name>group)» captures the match of «group» into the backreference “name”. you can use the two syntactic variations interchangeably.NET languages. Unfortunately. This does not work in PHP. You can use the pointy bracket flavor and the quoted flavors interchangeably. Currently. where single quotes may need to be escaped. PHP. starting with one. The PHP preg functions offer the same functionality. In a complex regular expression with many capturing groups.NET style: «(?<first>group)(?'second'group)». Named Capture with Python.NET framework also support named capture. As you can see. . Again. Use Round Brackets for Grouping All modern regular expression engines support capturing groups. PHP/preg.64 12. or one of the . the Microsoft developers decided to invent their own syntax. Simply use a name instead of a number between the curly braces. since they are based on PCRE. The numbers can then be used in backreferences to match the same text again in the regular expression. and number both kinds from left to right. Names and Numbers for Capturing Groups Here is where things get a bit ugly.NET's System. use «\k<name>» or «\k'name'». you can easily reference it by name. and will convert one flavor of named capture into the other when generating source code snippets for Python. rather than follow the one pioneered by Python. In PHP. you can reference the named group with the familiar dollar sign syntax: “${name}”. Here is an example with two capturing groups in .Text. To reference a capturing group inside the regex. When doing a search-and-replace. By assigning a name to a capturing group. RegexBuddy supports both Python's and Microsoft's style. Python's sub() function allows you to reference a named group as “\1” or “\g<name>”. The second syntax is preferable in ASP code. The first syntax is preferable in strings. PCRE and PHP Python's regex module was the first to offer a solution: named capture. You can reference the contents of the group with the numbered backreference «\1» or the named backreference «(?P=name)». Named Capture with . The open source PCRE library has followed Python's example. no other regex flavor supports Microsoft's version of named capture. which are numbered from left to right. you can use double-quoted string interpolation with the $regs parameter you passed to pcre_match(): “$regs['name']”.RegularExpressions The regular expression classes of the . where the sharp brackets are used for HTML tags. and the other using single quotes. The regex . and offers named capture using the same syntax. Python and PCRE treat named capturing groups just like unnamed capturing groups.

Easy and logical. you will get “abcd”. when using . Things are quite a bit more complicated with the . To keep things compatible across regex flavors. Probably not what you expected. in this case: three. continuing from the unnamed groups. However. if you do a search-and-replace with “$1$2$3$4” as the replacement. and reference them by name exclusively. The regex «(a)(?<x>b)(c)(?<y>d)» again matches „abcd”.NET framework. To make things simple.NET's regex support. since the regex engine does not need to keep track of their matches. If you do a search-and-replace with this regex and the replacement “\1\2\3\4”. from one till four. The . .65 «(a)(?P<x>b)(c)(?P<y>d)» matches „abcd” as expected. or make it non-capturing as in «(?:nocapture)». Non-capturing groups are more efficient. So the unnamed groups «(a)» and «(c)» get numbered first. Then the named groups «(?<x>b)» and «(?<y>d)» get their numbers. from left to right. but numbers them after all the unnamed groups have been numbered. just assume that named groups do not get numbered at all. you will get “acbd”. Either give a group a name. I strongly recommend that you do not mix named and unnamed capturing groups at all.NET framework does number named capturing groups from left to right. All four groups were numbered from left to right. starting at one.

The regex «(?i)te(?-i)st» should match „test” and „TEst”. but these differ widely. but not “teST” or “TEST”. E. Turning Modes On and Off for Only Part of The Regular Expression Modern regex flavors allow you to apply modifiers to only part of the regular expression. You have probably noticed the resemblance between the modifier span and the non-capturing group «(?:group)». while Pattern. you use a modifier span. /m enables "multi-line mode".CASE_INSENSITIVE) does the same in Java. In this mode. turns off single-line mode. the caret and dollar match before and after newlines in the subject string. Technically. «(?i)ignorecase(?-i)casesensitive(?i)ignorecase» is equivalent to «(?i)ignorecase(?i:casesensitive)ignorecase». precede each of their letters with a minus sign. in Perl. In this mode. Most programming languages allow you to pass option flags when constructing the regex object.g. m/regex/i turns on case insensitivity. (?i) turns on case insensitivity.g. Modifier Spans Instead of using two modifiers. It is obvious that the modifier span does not create a backreference.g. Regex Matching Modes All regular expression engines discussed in this tutorial support the following three matching modes: • • • /i makes the regex match case insensitive. you can add a mode modifier to the start of the regex. The latest versions of all tools and languages discussed in this book do. while (?ism) turns on all three options. E. To turn off several modes. If you insert the modifier (?ism) in the middle of the regex. the tool or language does not provide the ability to specify matching options. Pattern. E. Most tools that support regular expressions have checkboxes or similar controls that you can use to turn these modes on or off. You can turn off a mode by preceding it with a minus sign. the handy String. one to turn an option on. You can quickly test this. the non-capturing group is a modifier span that does not change any modifiers. .g. /s enables "single-line mode". and one to turn it off. no matter where you placed it. and turns on multi-line mode. Older regex flavors usually apply the option to the entire regular expression. Specifying Modes Inside The Regular Expression Sometimes.compile() does. In that situation. the modifier only applies to the part of the regex to the right of the modifier.66 13.matches() method in Java does not take a parameter for matching options like Pattern. Many regex flavors have additional modes or options that have single letter equivalents. (?i-sm) turns on case insensitivity. the dot matches newlines. Not all regex flavors support this.compile(“regex”. E.

”.10. again trying all possible combinations for the 9th. Reaching the end of the string again. The problem rears its ugly head when the 12th field does not start with a P. The next token is again the dot. When the 9th iteration consumes „9.10.5.){11}» had consumed „1. this leads to a catastrophic amount of backtracking.7. The dot matches a comma. Continuously failing. However.3. You can already see the root of the problem: the part of the regex (the dot) matching the contents of the field also matches the delimiter (the comma). so the dot continues until the 11th iteration of «.10.){11}P».11. A lazy quantifier will first repeat the token as few times as required.3.11.12.10.”. Greediness and laziness determine the order in which the regex engine tries the possible permutations of the regex pattern.*?.*?.”. It will backtrack to the point where «^(.e. the P checks if the 12th field indeed starts with P.12.4. the 10th iteration is expanded to „10. Since there is still no P. the engine backtracks to the 8th iteration.11.5. Because of the double repetition (star inside {11}).11. this regex looks like it should do the job just fine. This causes software like EditPad Pro to stop responding. this is exactly what will happen when the 12th field indeed starts with a P.6. and the {11} skips the first 11 fields. The customer was using the regexp «^(. It does not. The lazy dot and comma match a single comma-delimited field.8. the regex engine can no longer match the 11th iteration of «.”. But it does not give up there. the 10th could match just „11.9.4. But between each expansion. let's see why backtracking can lead to problems. or even crash as the regex engine runs out of memory trying to remember all backtracking positions.11”.6. and gradually expand the match as the engine backtracks through the regex to find an overall match. subsequently expanding it to „9.” as well as „11. First. The regex engine now checks whether the 13th field starts with a P. there are more possiblities to be tried. At first sight.12. In fact. it stopped responding) when trying to find lines in a comma-delimited text file where the 12th item on a line started with a “P”.8. Finally.». and gradually give up matches as the engine backtracks to find an overall match.*?. „9.*?.7. Let's say the string is “1. the same story starts with the 9th iteration. Atomic Grouping and Possessive Quantifiers When discussing the repetition operators or quantifiers.9.”. „9. Since there is no comma after the 13th field. the regex engine will backtrack. The dot matches the comma! However.10.10. 10th.67 14.2.”. giving up the last match of the comma. A greedy quantifier will first try to repeat the token as many times as possible.12. they do not change the fact that the regex engine will backtrack to try all possible permutations of the regular expression in case no match can be found. It backtracks to the 10th iteration. Catastrophic Backtracking Recently I got a complaint from a customer that EditPad Pro hung (i. expanding the match of the 10th iteration to „10. Because greediness and laziness change the order in which permutations are tried.”. and 11th iterations. the comma does not match the “1” in the 12th field. I explained the difference between greedy and lazy repetition.11.”.2. . they can change the overall regex match. You get the idea: the possible number of combinations that the regex engine will try for each line where the 12th field does not start with a P is huge.» has consumed „11.12.13”. At that point.

*?.4. In the above example. as do recent versions of PCRE and PHP's pgreg functions. you can be sure that the regex engine will try all those combinations. the engine has to backtrack to the regex token before the group (the caret in our example). Perl supports it starting with version 5. Python does not support atomic grouping. To make a quantifier possessive. If backtracking is required. Using atomic grouping. Tool and Language Support for Atomic Grouping and Possessive Quantifiers Atomic grouping is a recent addition to the regex scene. We want to match 11 commadelimited fields. forcing the regex engine to the previous one of the 11 iterations immediately.\r\n]*. Everything between (?>) is treated as one single token by the regex engine. the engine will still backtrack. . the solution is to be more exact about what we want to match. «x++» is the same as «(?>x+)».n}+».NET support atomic grouping. When nesting repetition operators.68 Preventing Catastrophic Backtracking The solution is simple.6. allowing the regex engine to fail faster. If repeating the inner loop 4 times and the outer loop 7 times results in the same overall match as repeating the inner loop 6 times and the outer loop 2 times. Possessive quantifiers are a limited form of atomic grouping with a cleaner notation. possessive quantifiers are only supported by the Java JDK 1. you should use atomic grouping to prevent the regex engine from backtracking. All versions of . as do all versions of RegexBuddy. no backtracking can take place once the regex engine has found a match for the group. without trying further options.0 and later. though the JDK documentation uses the term “independent group” rather than “atomic group”. and only supported by the latest versions of most regex flavors.){11})P». place a plus after it. Similarly. and PCRE version 4 and later. If the P cannot be found. But that is not always possible in such a straightforward manner.2. Note that you cannot make a lazy quantifier possessive. So the regex becomes: «^([^.\r\n]» is not able to expand beyond the comma. In our example. you can use «x*+». Atomic Grouping and Possessive Quantifiers Recent regex flavors have introduced two additional solutions to this problem: atomic grouping and possessive quantifiers. make absolutely sure that there is only one way to match the same match. once the regex engine leaves the group. the above regex becomes «^(?>(. In that case. the regex must retry the entire regex at the next position in the string.){11}P». The Java supports it starting with JDK version 1. The fields must not contain comma's. At this time. Their purpose is to prevent backtracking. But it will backtrack only 11 times. «x?+» and «x{m. and each time the «[^. Because the entire group is one token.4. The latest versions of EditPad Pro and PowerGREP support both atomic grouping and possessive quantifiers. we could easily reduce the amount of backtracking to a very low level by better specifying what we wanted. It would match the minimum number of matches and never expand the match because backtracking is not allowed. If there is no token before the group.

With simple repetition.4. With combined repetition. . The previous token is an atomic group. and is not immediately enclosed by an atomic group. That is.6.){11})P» is applied to “1. often it is not.5. the engine backtracks until the 6 can be matched. Again. you really should use atomic grouping and/or possessive quantifiers whenever possible.3. If the regex will be used in a tight loop in an application. If possessive quantifiers are available. or process huge amounts of data. you often can avoid the problem without atomic grouping as in the example above. troublesome regular expression.9. only failure. if you are smart about combined repetition. the regex engine did not cross the closing round bracket of the atomic group. the increase in speed is minimal.4.10. «P» failed to match. They do not speed up success. While «x[^x]*+x» and «x(?>[^x]*)x» fail faster than «x[^x]*x». the engine leaves the atomic group. using simple repetition. «\d++6» will not match at all. and just one attempt to match the atomic group. The most efficient regex for our problem at hand would be «^(?>((?>[^. When To Use Atomic Grouping or Possessive Quantifiers Atomic grouping and possessive quantifiers speed up failure by eliminating backtracking. The engine now tries to match the caret at the next position in the string.){11})P».9. you can reduce clutter by writing «^(?>([^. so the dot is initially skipped. you will not earn back the extra time to type in the characters for the atomic grouping.10. Now.){11})P».11. That's right: backtracking is allowed here. «\d+6» will match „123456” in “123456789”. and the match fails. then atomic grouping may make a difference. and declares failure. Still. In the latter case.8.\r\n]*). which fails. This shows again that understanding how the regex engine works on the inside will enable you to avoid many pitfalls and craft efficient regular expressions that match exactly what you want.3. so the engine backtracks.5.6.13”. so the engine backtracks to the dot. «\d+» will match the entire string. This fails. Failure is declared after 30 attempts to match the caret. rather than after 30 attempts to match the caret and a huge number of attempts to try all combinations of both quantifiers in the regex. If the final x in the regex cannot be matched. But the comma does not match “1”. When nesting quantifiers like in the above example. no backtracking is allowed. the cause of this is that the token «\d» that is repeated can also match the delimiter «6». The star is not possessive.7. Because the group is atomic. The engine walks through the string until the end. So far.*?. The caret matches at the start of the string and the engine enters the atomic group. That is what atomic grouping and possessive quantifiers are for: efficiency by disallowing backtracking. The star is lazy. the amount of time wasted increases exponentially and will very quickly exhaust the capabilities of your computer. Now comes the difference. and the comma matches too. so the group's entire match is discarded and the engine backtracks further to the caret. greedy repetition of the star is faster than a backtracking lazy dot. everything happened just like in the original.11.”. Sometimes this is desirable. The dot matches „1”.2.2. all backtracking information is discarded and the group is now considered a single token. «{11}» causes further repetition until the atomic group has matched „1.8. since possessive. The engine now tries to match «P» to the “1” in the 12th field. If you are simply doing a search in a text editor. With the former regex. the regex engine backtracks once for each character matched by the star.7.69 Atomic Grouping Inside The Regex Engine Let's see how «^(?>(.12.\r\n]*+. the amount of time wasted with pointless backtracking increases in a linear fashion to the length of the string. Note that atomic grouping and possessive quantifiers can alter the outcome of the regular expression match.

They are zero-width just like the start and end of line. but then give up the match and only return the result: match or no match. and begins matching the regex inside the lookahead. The other way around will not work. Let's try applying the same regex to “quit”. All regex flavors discussed in this book support lookaround. and discards the regex match. because the lookahead will already have discarded the regex match by the time the backreference is to be saved. The next token is the «u» inside the lookahead. the entire regex has matched. Lookarounds allow you to create regular expressions that are impossible to create without them. The exception is JavaScript. or that would get very longwinded without them. with the opening bracket followed by a question mark and an equals sign. Note that the lookahead itself does not create a backreference. That is why they are called “assertions”. The positive lookahead construct is a pair of round brackets. «q» matches „q”. I already explained why you cannot use a negated character class to match a “q” not followed by a “u”. So it is not included in the count towards numbering the backreferences. this will cause the engine to traverse the string until the „q” in the string is matched. I will explain why below. this means that the lookahead has successfully matched at the current position. Negative lookahead provides the solution: «q(?!u)». and „q” is returned as the match. They do not consume characters in the string. Collectively. The difference is that lookarounds will actually match characters. without making the u part of the match. These match. This does not match the void behind the string. The next token is the lookahead. If you want to store the match of the regex inside a backreference. with the opening bracket followed by a question mark and an explanation point. Because the lookahead is negative. If it contains capturing parentheses. So the next token is «u». we have the trivial regex «u». Positive lookahead works just the same. Positive and Negative Lookahead Negative lookahead is indispensable if you want to match something not followed by something else. The engine advances to the next character: “i”. As we already know. This causes the engine to step back in the string to “u”.70 15. Regex Engine Internals First. Inside the lookahead. Lookahead and Lookbehind Zero-Width Assertions Perl 5 introduced two very powerful constructs: “lookahead” and “lookbehind”. You can use any regular expression inside the lookahead. The engine takes note that it is inside a lookahead construct now. let's see how the engine applies «q(?!u)» to the string “Iraq”. you have to put capturing parentheses around the regex inside the lookahead. but only assert whether a match is possible or not. . The first token in the regex is the literal «q». «q(?=u)» matches a q that is followed by a u. (Note that this is not the case with lookbehind. which supports lookahead but not lookbehind. The next character is the “u”. However. When explaining character classes. like this: «(?=(regex))». At this point. The position in the string is now the void behind the string. it is done with the regex inside the lookahead. The negative lookahead construct is the pair of round brackets. and start and end of word anchors that I already explained. The engine notes success. the backreferences will be saved. The engine notes that the regex inside the lookahead failed. these are called “lookaround”. They are also called “zero-width assertions”.) Any valid regular expression can be used inside the lookahead.

Let's take one more look inside. «b» matches „b”. you could use «\b\w+(?<!s)\b». so the engine steps back from “i” in the string to “u”. More Regex Engine Internals Let's apply «(?<=a)b» to “thingamabob”. The next character is the second “a” in the string. so the engine continues with «i». which cannot match here. the “h”. “less than” symbol and an equals sign. Let's apply «q(?=u)i» to “quit”. so the positive lookbehind fails again. The next token is «b». The lookbehind continues to fail until the regex reaches the “m” in the string. So the lookbehind fails. But «i» cannot match “u”. the engine reports failure. It tells the regex engine to temporarily step backwards in the string. (Note that a negative lookbehind would have succeeded here. and put a token after it. The engine cannot step back one character because there are no characters before the “t”. The construct for positive lookbehind is «(?<=text)»: a pair of round brackets. To lookahead was successful. The engine again steps back one character. because there are no more q's in the string. The engine steps back and finds out that „a” satisfies the lookbehind. but will match the „b” (and only the „b”) in “bed” or “debt”. and the engine starts again at the next character. not only at the start. the successful match inside it causes the lookahead to fail. but does not match “bed” or “debt”. It matches one character: the first „b” in the string. the lookbehind tells the engine to step back one character. In this case. If you want to find a word not ending with an “s”. the engine temporarily steps back one character to check if an “a” can be found there.) Again. All remaining attempts will fail as well. Important Notes About Lookbehind The good news is that you can use lookbehind anywhere in the regex. with the opening bracket followed by a question mark. The positive lookbehind matches. the current position in the string remains at the “m”. and the entire regex has been matched successfully. This is definitely not the same as . Again. Because it is zero-width. the engine has to start again at the beginning. Again. The engine starts with the lookbehind and the first character in the string. So this match attempt fails. to make sure you understand the implications of the lookahead. Positive and Negative Lookbehind Lookbehind has the same effect. I have made the lookahead positive. and finds out that the “m” does not match «a». Since «q» cannot match anywhere else. «(?<!a)b» matches a “b” that is not preceded by an “a”. The engine steps back. and see if an “a” can be matched there. to check if the text inside the lookbehind can be matched there. «q» matches „q” and «u» matches „u”.71 Because the lookahead is negative. The next character is the first “b” in the string. but works backwards. «(?<=a)b» (positive lookbehind) matches the „b” (and only the „b”) in „cab”. using negative lookbehind. Negative lookbehind is written as «(?<!text)». using an exclamation point instead of an equals sign. It finds a “t”. the match from the lookahead must be discarded. It will not match “cab”. Since there are no other permutations of this regex. and notices that the „a” can be matched there.

the former will match „John” and the latter „John'” (including the apostrophe). The last regex. but only if all options in the alternation have the same length. . The string must be traversed from left to right. which works correctly. but you can use the question mark and the curly braces with the max parameter specified. EditPad Pro and PowerGREP. and \W in the character class). This includes PCRE. has a double negation (the \W in the negated character class). Some regex flavors support the above. Double negations tend to be confusing to humans. This means you can use literal text and character classes. The latter will also not match single-letter words like “a” or “I”. lookbehind is a valuable addition to the regular expression syntax. only allow fixed-length strings. but fixed lengths.NET framework. Personally. The only regex flavor that I know of that currently supports this is Sun's regex package in the JDK 1. (Hint: «\b» matches between the apostrophe and the “s”). However. the regular expression engine needs to be able to figure out how many steps to step back before checking the lookbehind. the semantics of applying a regular expression backwards are currently not well-defined. Finally. Until that happens. plus alternation with strings of different lengths. But each string in the alternation must still be of fixed length. Microsoft has promised to resolve this in version 2. some more advanced flavors support the above. The bad news is that you cannot use just any regex inside a lookbehind. The reason is that regular expressions do not work backwards.72 «\b\w+[^s]\b». This means you can still not use the star or plus. including those used by Perl 5 and Python. including infinite repetition. I will leave it up to you to figure out why. Even with these limitations. I find the lookbehind easier to understand. many regex flavors.4. Therefore. You cannot use repetition or optional items. and will allow you to use any regex. though. You can use alternation. Therefore. JavaScript does not support lookbehind at all. Technically. RegexBuddy. PHP. You can use any regex of which the length of the match can be predetermined.0 of the . the . The correct regex without using lookbehind is «\b\w*[^s\W]\b» (star instead of plus. alternation and character classes inside lookbehind. inside lookbehind.NET framework can apply regular expressions backwards. Finally. plus finite repetition. These regex flavors recognize the fact that finite repetition can be rewritten as an alternation of strings with different. When applied to “John's”. Not to regex engines. I recommend you use only fixed-length strings. so only literals and character classes can be used.

Matching a word containing “cat” is equally easy: «\b\w*cat\w*\b». But this method gets unwieldy if you want to find any word between 6 and 12 letters long containing either “cat”. I would like to give you another. Combining the two. we want a word that is 6 letters long. If «cat» can be successfully matched. If not. The confusing part is that the lookaround is zero-width. Unfortunately. in the 6letter word. reducing the number of characters matched by «\w*». Easy enough. Second.73 16. the current position in the string is still at the beginning of the 6-letter word. The engine will then backtrack. . Let's say we want to find a word that is six letters long and contains the three subsequent letters “cat”. is a very powerful concept. the last «\b» in the regex is guaranteed to match where the second «\b» inside the lookahead matched. “dog” or “mouse”. then the regex will traverse part of the string twice. After that. it is often underused by people new to regular expressions. until «cat» can be matched. and therefore the lookahead. causing the engine to advance character by character until the next 6-letter word. Easy! Here's how this works. Matching a 6-letter word is easy with «\b\w{6}\b». Testing The Same Part of The String for More Than One Requirement Lookaround. Because we already know that a 6-letter word can be matched at the current position. matches only when the current character position in the string is at the start of a 6-letter word in the string. This sub-regex. The lookahead is zero-width. a bit more practical example. At this position will the regex engine attempt the remainder of the regex. We just specify all the options and hump them together using alternation: «cat\w{3}|\wcat\w{2}|\w{2}cat\w|\w{3}cat». the lookahead will fail. we get: «(?=\b\w{6}\b)\b\w*cat\w*\b». To make this clear. At each character position in the string where the regex is attempted. we can match this without lookaround. If «cat» cannot be matched. the word we found must contain the word “cat”. we know that «\b» matches and that the first «\w*» will match 6 times. First. or a lookbehind is preceded by another piece of regex. if any. because lookaround is a bit confusing. So when the regex inside the lookahead has found the 6-letter word. which I introduced in detail in the previous topic. So if you have a regex in which a lookahead is followed by another piece of regex. Our double-requirement-regex has matched successfully. the engine will first attempt the regex inside the positive lookahead. at the next character position in the string. Lookaround to The Rescue In this example. where the lookahead will fail. the engine has no other choice but to restart at the beginning of the regex. and the engine will continue trying the regex from the start at the next character position in the string. the second «\w*» will consume the remaining letters. This is at the second letter in the 6-letter word we just found. we basically have two requirements for a successful match. Actually.

But optimizing things is a good idea if this regex will be used repeatedly and/or on large chunks of data in an application you are developing. instead of the entire word. Since it is zero-width itself.74 Optimizing Our Solution While the above regex works just fine. up to and including “cat”. One last. Remember that the lookahead discards its match.9}(cat|dog|mouse)\w*». . we can remove them. This is not a problem if you are just doing a search in a text editor. and even at one character beyond the 6-letter word. But we can optimize the first «\w*». once you get the hang of it. A More Complex Problem So. as I did above. If we omitted the «\w*». I said the third and last «\b» are guaranteed to match. Very easy. “dog” or “mouse”? Again we have two requirements. there can never be more than 3 letters before “cat”. So we have «(?=\b\w{6}\b)\w{0. it would still cause the regex engine to try matching “cat” at the last two letters. it will match 6 letters and then backtrack. but if a 6-letter word does not contain “cat”. which we can easily combine using a lookahead: « \b(?=\w{6.3}cat\w*». “dog” or “mouse” into the first backreference. at the last single letter. leaving: «(?=\b\w{6}\b)\w*cat\w*». the resulting match would be the start of a 6-letter word containing “cat”. So the final regex is: «\b(?=\w{6}\b)\w{0. we cannot remove it because it adds characters to the regex match. what would you use to find any word between 6 and 12 letters long containing either “cat”. Note that making the asterisk lazy would not have optimized this sufficiently. it is not the most optimal solution.12}\b)\w{0. The lazy asterisk would find a successful match sooner. optimization involves the first «\b».3}cat\w*». You can discover these optimizations by yourself if you carefully examine the regex and follow how the regex engine applies it. so it does not contribute to the match returned by the regex engine. and therefore does not change the result returned by the regex engine. As it stands. there's no need to put it inside the lookahead. This regex will also put “cat”. So we can optimize this to «\w{0. minor. But we know that in a successful match. Though the last «\w*» is also guaranteed to match. Since it is zero-width.3}».

To keep things simple. The dot and negative lookahead match any character that is not the first character of the start of a section.)*?stop)». it will continue after „stop”. and «stop» as the regex matching the end of the section. When we apply the regex again to the same string or file.)*?stop». Substitute «wanted». This is possible with lookahead. at which point stop cannot be matched and thus the regex will fail. The final regular expression will be in the form of «wanted(?=insidesection)». Because of the negative lookahead inside the star. and «start» as the regex matching the start of the section. but only inside title tags. Note that these two rules will only yield success if the string or file searched through is properly translated into sections. The reason is that this regular expression consumes the entire section. So we have to resort to using lookahead only. That is. If we could.*?stop)». we found a match before a section rather than inside a section. Effectively. The final regular expression becomes: «wanted(?=((?!start). «start. This. Finding Matches Only Inside a Section of The String Lookahead allows you to create regular expression patterns that are impossible to create without it. I used a lazy star to make the regex more efficient. Second.*?». How do we know if we matched «wanted» inside a section? First. The star is obviously not of fixed length. First we match the string we want. the star will also stop at the start of a section. If “wanted” occurs only once inside the section. and end with the section stop. «start» and «stop» with the regexes of your choice. So inside the lookahead we need to look for a series of unspecified characters that do not match the start of a section anywhere in the series. The regex engine will refuse to compile this regular expression. If not. Lookbehind must be of fixed length. this will not work. we must not be able to match «start» between matching «wanted» and matching «stop». we repeat zero or more times with the star. replacing a certain word with another. we must be able to match «stop» after matching «wanted». this will not work if “wanted” occurs more than once inside a single section. After this. A title tag starts with «<H[1-6]» and . However. you can easily build a regex to do a search and replace on HTML files.75 17. not after „wanted”. You may be tempted to use a combination of lookbehind and lookahead like in «(?<=start. we need to match the end of the section. we need to use «. One example is matching a particular regex only inside specific sections of the string or file searched through. Since we do not know in advance how many characters there will be between “start” and “wanted”. the lazy star will continue to repeat until the end of the section is reached. we can do without lookahead. because lookahead is zero-width. each match of «start» must be followed exactly by one match of «stop». and then we test if it is inside the proper section. Example: Search and Replace within Header Tags Using the above generic regular expression. this is written as: «((?!start). So we need a way to match „wanted” without matching the rest of the section. In a regex. However. The entire section is included in the regex match.*?stop» would do the trick. I will use «wanted» as a substitute for the regular expression that we are trying to match inside the section.*?wanted.*?)wanted(?=. we found a match after a section rather than inside a section.

But lookahead is what we need here. .)*?</H[1-6]>)». I omitted the closing > in the start tag to allow for attributes. or negative lookbehind. You may have noticed that I escaped the < of the opening tag in the final regex. Escaping the < takes care of the problem.76 ends with «</H[1-6]>». I did that because some regex flavors interpret «(?!<» as identical to «(?<!». So the regex becomes «wanted(?=((?!\<H[1-6]).

«\G» matches at the start of the match attempt. the position where the last match ended is a “magical” value that is remembered separately for each string variable. During the fifth attempt.etc.77 18.. Applying «\G\w» to the string “test string” matches „t”. rather than the end of the previous match. Applying it again matches „e”. «\G» matches at the start of the string in the way «\A» does. During the first match attempt. When a match is found.. Continuing at The End of The Previous Match The anchor «\G» matches at the position where the previous match ended. the stored position for «\G» is reset to the start of the string. The regex in the while loop searches for the tag's opening bracket. you could parse an HTML file in the following fashion: while ($string =~ m/</g) { if ($string =~ m/\GB>/c) { } elsif ($string =~ m/\GI>/c) { } else { } } # Bold # Italics # . End of The Previous Match vs Start of The Match Attempt With some regex flavors or tools. and the regexes inside the loop check which tag we found. If a match attempt fails. without having to write a single big regex that matches all tags you are interested in. The result is that «\G» matches at the end of the previous match result only when you do not move the text cursor between two searches. . \G Magic with Perl In Perl. specify the continuation modifier /c. All in all.. All this is very useful to make several regular expressions work together. the only place in the string where «\G» matches is after the second t. But that position is not followed by a word character.. This way you can parse the tags in the file in the order they appear in the file. and move the text cursor to the end of the match.g. EditPad Pro will select the match. This is the case with EditPad Pro. This means that you can use «\G» to make a regex continue in a subject string where another regex left off. E. To avoid this. The position is not associated with any regular expression. The fifth attempt fails. rather than at the end of the previous match result. so the match fails. this makes a lot of sense in the context of a text editor. where «\G» matches at the position of the text cursor. The 3rd attempt yields „s” and the 4th attempt matches the second „t” in the string.

E. What you can do though is to add a line of code to make the match attempt of the second Matcher start where the match of the first Matcher ended.78 \G in Other Programming Langauges This flexibility is not available with most other programming languages. in Java. «\G» will then match at this position. The Matcher is strictly associated with a single regular expression and a single subject string. the position for «\G» is remembered by the Matcher object.g. .

Otherwise. like in «(?(?=condition)(then1|then2|then3)|(else1|else2|else3))». the syntax becomes «(?(?=regex)then|else)». you can use any regular expression. there is no need to use parentheses around the then and else parts. and the vertical bar with it. If you want to use alternation. Remember that the lookaround constructs do not consume any characters. You may omit the else part. . you can use the lookahead and lookbehind constructs. then the regex engine will attempt to match the then part. The syntax consists of a pair of round brackets. If the if part evaluates to true. you will have to group the then or else together using parentheses.79 19. then the regex engine will attempt to match the then or else part (depending on the outcome of the lookahead) at the same position where the if was attempted. For the then and else. If-Then-Else Conditionals in Regular Expressions A special construct «(?ifthen|else)» allows you to create conditional regular expressions. The opening bracket must be followed by a question mark. immediately followed by the if part. This part can be followed by a vertical bar and the else part. If you use a lookahead as the if part. Because the lookahead has its own parentheses. For the if part. Using positive lookahead. immediately followed by the then part. the else part is attempted instead. Otherwise. the if and then parts are clearly separated.

That makes the comments really stand out.](?#month)(0[1-9]|1[012])[. Some software.g. . such as RegexBuddy. E. Adding Comments to Regular Expressions If you have worked through the entire tutorial.80 20. Therefore. I guess you will agree that regular expressions can quickly become rather cryptic. many modern regex flavors allow you to insert comments into regexes. I could clarify the regex to match a valid date by writing it as «(?#year)(19|20)\d\d[/. The regex engine ignores everything after the «(?#» until the first closing round bracket./. Now it is instantly obvious that this regex matches a date in yyyy-mm-dd format. EditPad Pro and PowerGREP can apply syntax coloring to regular expressions while you write them. as long as it does not contain a closing round bracket.](?#day)(0[1-9]|[12][0-9]|3[01])». enabling the right comment in the right spot to make a complex regular expression much easier to understand. The syntax is «(?#comment)» where “comment” is be whatever you want.